Complete Mastery of AI Harness Engineering for Everyone
Are you explaining the same task to AI from scratch every time? A practicing developer shares the complete setup they refined over 7 months.—Even as a non-developer, build the harness with just Claude Code and actually hand one of your repetitive tasks over to AI.
My own automation skill that executes a weekly recurring task with just "one word".
Onboarding documents and a memory folder where AI remembers the rules, formats, and context of my work (AI setup dedicated to my work)
The planning skill that turns the vague “I want to do this” into an actionable plan
A work environment where calendars, email, and spreadsheets are connected to AI, with clear safety boundaries defining how much you can delegate to it
Practical automation experience you can put to use today, including organizing folders, reconciling receipts, and writing meeting minutes.
A safety net (Git) that lets you undo anything AI does at any time, and an eye for validating the results
This course does not teach the skills of a specific job. It covers the essentials needed to properly hand off any task to AI. Whether you’re a marketer, content creator, planner, or developer—the way to make AI not just a tool but an assistant who knows your work is the same. That’s why the title is "Essential for Everyone."
Harness—a term derived from the harness used to connect horses to carriages, referring to the entire setup that connects AI to real-world work. This includes documents that know your work, accumulated memory, manuals for repetitive tasks, tool integrations and permissions, and verification mechanisms. It is already an established concept among developers overseas, and this is the first course in Korea to teach it in the language of office workers who don't know how to code.
🙋 Have you ever had an experience like this?
The experience of asking AI to do a task and getting a plausible result, but being unable to use it as is
Experiencing having to give last week’s explanation all over again from the beginning this week
"Write the conclusion first," "I told you, that’s not the format" — the experience of repeating the same nagging every time
The problem is neither the AI’s abilities nor your prompts. It’s that all the rules of your work exist only in your head and have never been turned into a file. This course solves that.
🔧 It’s not a prompt—it’s a “setup”
A well-written prompt solves the day, but a setup keeps working tomorrow too. In this course, you assemble six components one by one—and when you’re done, AI becomes not "a new hire who arrives every day," but an assistant who knows my work.
💎 4 Reasons This Course Is Different
1️⃣ The result is not ‘knowledge’ but ‘automation that works’
The completion standard is different—not “I learned it well,” but whether one of my repetitive tasks has actually been handed over to AI. When the course ends, the results of my computer’s work on my behalf will be waiting for me every morning.
2️⃣ I’ll bring you the exact workflow used by a working developer.
The instructor is a working iOS developer who works with AI every day at the company. Rather than hastily created examples for the course, we bring you the exact setups and procedures currently being used in real-world work. Only proven methods remain—the instructor has already gone through the trial and error.
3️⃣ Knowledge that remains relevant even when tools change and versions are upgraded
How to plan, write a requirements document, break down large tasks into smaller ones, and validate the results—these are core concepts for working with harnesses. Even if Claude Code is replaced by another tool, this remains essential knowledge, no matter how much LLM versions advance. Tool usage has a shelf life, but how to get work done remains.
It starts with a work plan. Starting from the vague thought, "I want to automate this," you go back and forth with AI to complete a requirements document—just answer the questions, and the plan takes shape. You can see this scene firsthand in the free public class.
4️⃣ Non-developers can also use 'Git' — the reassurance that you can undo changes
To feel comfortable entrusting tasks to AI, you need a safety net that lets you undo things even if you mess up. We teach you how to use Git, the version control tool developers use, without knowing a single line of code. We show you how to create save points and return to them at any time—with this safety net, you can entrust even major tasks to AI. Few courses properly teach non-developers this skill.
🗺 Journey
Free section (90 minutes): Installation → first successful automation (organizing the Downloads folder and analyzing disk space) → file and Git fundamentals + a demonstration lesson on task planning are all free. Try it first, then decide.
Paid section: Work planning → Harness assembly (connecting documents, memory, skills, tools, permissions, and verification) → Hands-on migration практиce
👍 Recommended for:
All working professionals who want to hand repetitive tasks over to AI and free up time — marketers, content creators, planners, operations professionals, self-employed business owners, and even developers who want to learn how to set everything up properly
People who have used ChatGPT but have never actually turned it into an AI agent
If you’re curious about vibe coding but want to use it not for “building apps,” but for the work you actually need
🙅 This is not suitable for the following people
Developers who only want advanced development (code reviews and architecture)
Those expecting to “monetize a certain amount per month” — this course promises only to actually hand over my repetitive tasks.
✅ Please check this before enrolling.
A paid Claude subscription is required (Practice tool: Claude Code only)
The course screen is based on macOS — for Windows, supplementary guides for sections that differ are provided in documentation (real-device testing is only available on macOS).
The hands-on section will be added in a post-release update (included with your course access — no additional cost).
👨💻 Instructor Introduction — Dumbelloper
An iOS developer with 8 years of industry experience. After discovering AI two years ago, I spent the past 7 months researching harness engineering—focusing on configuring AI as reliable agents.
Even outside the company, I run thread automation pipelines, a YouTube channel, and personal records with the same setup — whether it’s work or a hobby, the setup can be replicated as-is
I run 6 courses on Inflearn, with a total of 525 students — I’ve taught before.
Recommended for these people
Who is this course right for?
An office worker who does the same reports, reconciliations, and data organization every week, yet has to explain everything to AI from scratch each time
Those who have tried AI but thought, "What’s so convenient about this?"
Those who end up checking everything themselves from the beginning because they don’t trust AI-generated results.
A non-developer curious about Claude Code, which developers use, but too intimidated by the terminal to get started.
For those who clearly have tasks they want to automate but don’t know what to ask for.
Need to know before starting?
No coding knowledge is required at all
Claude paid subscription & hands-on environment are based on macOS, and supplementary guides are provided for differences in installation and some commands so that the course can also be taken on Windows.